JP2016048423A - Feedback control device and electrically-driven power steering apparatus - Google Patents

Abstract

PROBLEM TO BE SOLVED: To achieve a feedback control device capable of continuing a control action safely even when a soft error occurs, and to achieve an electrically-driven power steering apparatus.SOLUTION: A controller 202 determines a d-axis target voltage Vd and a q-axis target voltage Vq which are control values for controlling an output value of the electrically-driven power steering apparatus, on the basis of a d-axis error signal δId and a q-axis error signal δIq which are input values. In this processing, for a period of time before a soft error occurs, the controller 202 determines the d-axis target voltage Vd and the q-axis target voltage Vq by PID control where current input values and past input values are used. On the other hand, for a period of time after the soft error occurs, the controller determines the d-axis target voltage Vd and the q-axis target voltage Vq by P control where the current input values are used and the past input values are not used. After this period of time has passed, the PID control is resumed.SELECTED DRAWING: Figure 1

Description

  The present invention relates to a feedback control device and an electric power steering device using the feedback control device.

  In recent years, as control automation using an electronic control device configured using a microcomputer or the like proceeds, demands for safety and reliability of the electronic control device have increased. For example, when an abnormality occurs in the electronic control device, it is required to immediately stop the operation or continue the control operation as much as possible while ensuring safety.

  As an example of performing control using the electronic control device as described above, an electric power steering device for a vehicle is known. With recent performance improvements, electric power steering devices are also applied to heavy vehicles with large weights. However, especially in large vehicles, if the operation of the electric power steering device is stopped, a large amount of manual steering force is required. Necessary. Therefore, it is necessary to continue the operation of the electric power steering apparatus while ensuring safety even when a failure occurs.

  On the other hand, with the recent increase in the degree of integration of microcomputers, there is a concern about an increase in the occurrence rate of soft errors due to disturbances such as cosmic rays. The soft error is also called SEU (Single Event Upset), and is classified as a transient fault (transient fault) that occurs during the operation of the microcomputer. When a soft error occurs, abnormal data is generated because bit inversion occurs in data stored in the microcomputer. In this case, if the data is stored in the memory of the microcomputer, abnormal data can be detected and corrected by an error correction code (ECC). However, since the error correction code is not usually added to the data stored in the register in the microcomputer processor, the register data error due to a soft error may cause the processor to malfunction or run away. is there. Further, in this case, data stored in the memory may be lost due to malfunction or runaway of the processor. As a result, it becomes difficult to continue control safely.

  In the electronic control device, it is preferable to continue the control safely even when the data abnormality due to the soft error as described above occurs. In this regard, in Patent Document 1 below, in the electric power steering apparatus that performs feedback control based on the motor current detected by the current detector, when the current detector fails and feedback control becomes impossible, It is disclosed that the steering assist is continued by switching to feedforward control that does not use the motor current detected by the current detector.

JP 2006-44338 A

  In the technique described in Patent Document 1, when switching from feedback control to feedforward control, a control error with respect to a target value becomes too large. Moreover, after switching to feedforward control once, it cannot return to feedback control. Therefore, it is not appropriate for dealing with transient failures such as soft errors.

  The present invention has been made to solve the above-described problems of the prior art. A main object of the present invention is to realize a feedback control device and an electric power steering device that can continue a control operation safely even when a soft error occurs.

A feedback control device according to an aspect of the present invention determines a control value for controlling the output value based on an input value corresponding to a difference between a target value and an output value, and before a soft error occurs. In the first period, the control value is determined by the first feedback control using the current input value and the past input value, and in the predetermined second period after the soft error occurs, The control value is determined by the second feedback control that does not use the past input value using the input value, and the first feedback control is resumed in the third period after the second period has elapsed. To do.
A feedback control device according to another aspect of the present invention determines a control value for controlling the output value based on an input value corresponding to a difference between a target value and an output value, and a soft error occurs. In a first period before the steady-state deviation between the target value and the output value is a predetermined first value, and in a predetermined second period after the soft error occurs, the target The steady-state deviation between the value and the output value is greater than the first value.
An electric power steering apparatus according to the present invention includes the above feedback control apparatus, a motor that generates an assist torque for the steering mechanism, and an inverter that drives the motor based on a control value determined by the feedback control apparatus.

  According to the present invention, it is possible to realize a feedback control device and an electric power steering device that can continue a control operation safely even when a soft error occurs.

It is a figure which shows the structure of the electric power steering apparatus which concerns on the 1st Embodiment of this invention. It is a figure which shows an example of the hardware constitutions of a control system. It is a figure showing an example of processing at the time of soft error occurrence by the 1st processing method. It is a figure which shows the example of the process at the time of the soft error occurrence by a 2nd processing method. It is a figure which shows the example of the process at the time of the soft error occurrence by the 3rd processing method. It is the figure which showed the mode of the change of the proportional gain coefficient Kp in the 3rd processing method, the integral gain coefficient Ki, and the differential gain coefficient Kd. It is the figure which showed the mode of the change of the internal data of the integrator in a 3rd processing method, and a differentiator. It is a figure which shows the mode of a change of the output value of a control system in case a control target value converts into rectangular wave shape. It is a figure which shows the mode of a change of the output value of a control system when a control target value changes to a lamp input state. It is a figure which shows the whole structure of the electric power steering apparatus which concerns on the 2nd Embodiment of this invention. It is a figure which shows the control block of the controller which concerns on the 3rd Embodiment of this invention. It is the figure which showed the mode of the change of the coefficient matrix in the 3rd Embodiment of this invention. It is the figure which showed the mode of the change of the internal data of the integrator in the 3rd Embodiment of this invention.

  Hereinafter, embodiments of the present invention will be described with reference to the drawings.

(First embodiment)
FIG. 1 is a diagram illustrating a configuration of an electric power steering apparatus according to a first embodiment of the present invention. FIG. 1A is a diagram showing the overall configuration of the electric power steering apparatus, and FIG. 1B is a control block diagram of the controller 202 in the electric power steering apparatus.

  As shown in FIG. 1A, the electric power steering apparatus includes a motor 2, an inverter 100, and a control system 200. The motor 2 is a three-phase AC motor for generating an assist torque for a steering mechanism (not shown) provided in a vehicle on which the electric power steering device is mounted, and its drive is controlled by the inverter 100. The inverter 100 PWMs the switching elements of each phase (not shown) based on the target duty cycles Dual, Dval and Dwall output from the control system 200 to the U phase, V phase and W phase of the motor 2. Each is driven by control. As a result, the currents Iu, Iv, and Iw of each phase flowing through the motor 2 are controlled, and the motor 2 is driven.

  The control system 200 functionally includes a current command value calculation unit 201, a controller 202, a two-phase three-phase conversion unit 203, a duty calculation unit 204, and a three-phase two-phase conversion unit 205.

  The currents Iu, Iv, and Iw flowing through the motor 2 are detected by a current sensor 110 provided on each phase AC output line connected between the inverter 100 and the motor 2 and input to the control system 200. . The three-phase / two-phase conversion unit 205 converts the values of the currents Iu, Iv, and Iw detected by the current sensor 110 into a d-axis current Id and a q-axis current Iq.

  The current command value calculation unit 201 calculates a d-axis current command value Id * and a q-axis current command value Iq * based on a torque command input from the outside according to the steering amount of the driver with respect to the steering mechanism. A value obtained by subtracting the d-axis current Id and the q-axis current Iq obtained by the three-phase / two-phase converter 205 from the d-axis current command value Id * and the q-axis current command value Iq * is a d-axis error signal δId. And q-axis error signal δIq is input to controller 202. That is, the d-axis error signal δId and the q-axis error signal δIq that are input values to the controller 202 are a torque command that is a control target value of the electric power steering device and an output of the motor 2 that is an output value with respect to this torque command. It is determined according to the difference with the torque.

  The controller 202 functions as a feedback control device according to an embodiment of the present invention, and performs feedback control according to an input value. Specifically, the d-axis target voltage Vd and the q-axis target voltage are used as control values for controlling the output torque of the motor 2 in response to the torque command based on the input d-axis error signal δId and q-axis error signal δIq. Vq is determined. This feedback control will be described in detail later with reference to FIG.

  The two-phase / three-phase conversion unit 203 is configured such that the d-axis target voltage Vd and the q-axis target voltage determined by the controller 202 based on magnetic pole position information output from a magnetic pole position sensor (not shown) provided in the motor 2. Vq is converted into voltage target values Vu, Vv, and Vw for each phase.

  Based on the voltage target values Vu, Vv, and Vw obtained by the two-phase / three-phase converter 203, the duty calculator 204 calculates target duty cycles Double, Dval, and Dwall for each phase, and outputs them to the inverter 100.

  Next, a control block diagram of the controller 202 shown in FIG. As shown in FIG. 1B, the controller 202 has control blocks of an integrator 2021, a differentiator 2022, amplifiers 2023, 2024 and 2025, and an adder 2026.

  The amplifier 2023 multiplies the values of the d-axis error signal δId and q-axis error signal δIq input to the controller 202 as input values by a predetermined proportional gain coefficient Kp, and outputs the calculation result to the adder 2026.

  The integrator 2021 holds the results of accumulating the values of the d-axis error signal δId and the q-axis error signal δIq input to the controller 202 in time series as internal data, so that the d-axis error signal δId and q An integral value of the axis error signal δIq is calculated. The amplifier 2024 multiplies the integral values of the d-axis error signal δId and q-axis error signal δIq obtained by the integrator 2021 by a predetermined integral gain coefficient Ki, and outputs the calculation result to the adder 2026.

  The differentiator 2022 calculates the values of the current d-axis error signal δId and q-axis error signal δIq input to the controller 202 and the immediately preceding d-axis error signal δId and q-axis error signal δIq held as internal data. By calculating the difference from the value, the differential values of the d-axis error signal δId and the q-axis error signal δIq are calculated. The amplifier 2025 multiplies the differential values of the d-axis error signal δId and q-axis error signal δIq obtained by the differentiator 2022 by a predetermined differential gain coefficient Kd, and outputs the calculation result to the adder 2026.

  The adder 2026 determines the d-axis target voltage Vd and the q-axis target voltage Vq as control values based on the sum of the above-described calculation values.

  Note that the feedback control of the controller 202 performed by the control block diagram as described above is called PID control. Hereinafter, the PID control performed by the controller 202 will be described in more detail.

  When the input value in PID control is expressed as u (t) and the output control value is expressed as y (t), the relationship between the input value u (t) and the control value y (t) is the proportional gain coefficient Kp, Using the integral gain coefficient Ki and the differential gain coefficient Kd, the following expression (1) is used. The input value u (t) represents a control deviation (deviation between the output value and the target value).

  In the right side of the above equation (1), the value of the first term is determined according to the input value u (t) at the current time t. On the other hand, the value of the second term, which is the product of the integral gain coefficient Ki and the integral value calculation result by the integrator 2021, is determined according to the input value accumulation result from time 0 to the current time t. That is, it can be said that the value of the second term is history data determined from the history of past input values. The value of the third term, which is the product of the differential gain coefficient Kd and the differential value calculation result by the differentiator 2022, is determined according to the change in the current input value u (t) with respect to the previous input value. That is, the value of the third term can also be said to be history data determined from the history of past input values, similar to the value of the second term, although the accumulation time is short.

  As described above, the control value y (t) by PID control is obtained from the history data of the current input value u (t) and past input values. That is, PID control is feedback control using a current input value and a past input value.

  Next, the hardware configuration of the control system 200 will be described. FIG. 2 is a diagram illustrating an example of a hardware configuration of the control system 200. Each configuration of the control system 200 shown in FIG. 1A can be realized by, for example, a hardware configuration as shown in FIG.

  In the example illustrated in FIG. 2, the control system 200 includes CPUs 210-1 and 210-2, a comparator 211, an ECC circuit 212, a memory 213, and an I / O device 214. The CPUs 210-1 and 210-2 execute the programs stored in the memory 213, thereby realizing control processes corresponding to the respective control blocks in FIG. In the control system 200, redundant processing is performed by operating the two CPUs in parallel.

  The comparator 211 compares the output of the CPU 210-1 with the output of the CPU 210-2 to determine whether there is an abnormality in these processing results. When a permanent failure such as a hardware failure or a transient failure such as a soft error occurs in the CPU 210-1 or CPU 210-2, the outputs do not match between the CPU 210-1 and the CPU 210-2. In this case, the comparator 211 determines that one of the processing results is abnormal.

  When an abnormality is determined by the comparator 211, the CPUs 210-1 and 210-2 each execute a predetermined reset process. In this reset process, variables necessary for resuming the control process, such as variables used by the operating system such as various pointers, semaphores, counters, and program counters are recovered and reset, and stored in the memory 213. Reset various control data to initial values. This control data includes an integral value for PID control obtained by the integrator 2021 in FIG. 1B and a differential value for PID control obtained by the differentiator 2022. It should be noted that due to recent improvements in processor operating speed, the resetting of variables as described above can be performed in a time that is negligible compared to the time constant of the control target.

  When the reset process ends, the CPUs 210-1 and 210-2 resume the control process. In addition, when abnormality determination continues after restarting control processing due to a permanent failure such as a hardware failure, it is preferable to stop the operation of the control system 200. In order to determine this, hardware diagnosis or the like may be performed after the reset process.

  The ECC circuit 212 adds an error correction code (ECC) to data input / output between the CPUs 210-1 and 210-2 and the memory 213, and when there is an error in the data based on the error correction code. Detect and correct the error. The I / O device 214 performs data input / output between the control system 200 and an external device.

  The hardware configuration of the control system 200 described above is merely an example, and other hardware configurations may be used. For example, processing may be made redundant using three or more CPUs, or only one CPU may be used without redundancy. Any hardware configuration may be used as long as the control system 200 as shown in FIG. 1A can be realized.

  Next, processing when a soft error occurs in the controller 202 will be described. When a soft error occurs in the CPU 210-1 or CPU 210-2 in the controller 202, the comparator 211 makes an abnormality determination as described above, and in response to this, the CPU 210-1 and the CPU 210-2 perform a reset process. Is called. As a result, in the control block shown in FIG. 1B, the contents of the integrator 2021 and the differentiator 2022 are reset and their internal data are erased. PID control using an integral value or a differential value cannot be executed.

  Therefore, the controller 202 performs feedback control using the current input value without using the past input value within a predetermined period after restarting the control processing, and performs integration by the convergence operation in this feedback control. The internal data of the differentiator 2021 and the differentiator 2022 are recovered. Specifically, in the above equation (1), by setting the integral gain coefficient Ki of the second term and the differential gain coefficient Kd of the third term to 0, only the value of the first term corresponding to the current input value is obtained. The control value is obtained by P control, which is feedback control using. At this time, the d-axis error signal δId and the q-axis error signal δIq that are input values are continuously input to the integrator 2021 and the differentiator 2022, thereby recovering these internal data. When the internal data of the integrator 2021 and the differentiator 2022 can be recovered, the controller 202 resumes PID control by returning the integral gain coefficient Ki and the differential gain coefficient Kd to their original values, and returns to the normal control operation. .

  A series of processes when the soft error described above will be described in more detail with reference to FIGS. 3 is a diagram illustrating an example of processing when a soft error occurs according to the first processing method, and FIG. 4 is a diagram illustrating an example of processing when a soft error occurs according to the second processing method. These are figures which show the example of the process at the time of the soft error occurrence by a 3rd processing method. When a soft error occurs, the controller 202 can perform processing according to any of the processing methods shown in these drawings. In any of FIGS. 3 to 5, it is assumed that the controller 202 operates normally and performs a control operation based on PID control before a soft error occurs at time T0.

  In the first processing method shown in FIG. 3, when a soft error occurs at time T0, the controller 202 performs control processing after time T1 after performing the above-described reset processing, as PID control before the occurrence of soft error. To P control which is feedback control without using history data. Thereafter, during the period from time T1 to time T2, the internal data of the integrator 2021 and the differentiator 2022 necessary for PID control is recovered by performing a convergence operation. When the internal data of the integrator 2021 and the differentiator 2022 can be recovered in this way, the controller 202 switches the control processing after the time T2 from P control to PID control, and returns to the control state before the occurrence of the soft error.

  Note that the internal data of the differentiator 2022 is the values of the immediately preceding d-axis error signal δId and q-axis error signal δIq as described above, so that the recovery is achieved when at least one control cycle (control frame) has passed. Is possible. Therefore, it is preferable to preferentially recover the internal data of the differentiator 2022 during the period from the time T1 to the time T3, and subsequently recover the internal data of the integrator 2021 during the period from the time T3 to the time T2. At this time, in the period from time T3 to time T2, P control (PD control) in consideration of the differential value obtained by the differentiator 2022 may be performed.

  In the second processing method shown in FIG. 4, when a soft error occurs at time T0, the controller 202 switches the control processing after time T1 from PID control to P control, as in the first processing method. At this time, in the period from time T1 to time T4, by reducing the control rate of the controller 202, the period of P control performed in this period is made longer than the period of PID control before time T1. As a result, the controller 202 reduces the ratio of the time allocated to the control process from the normal time, and performs the processes necessary to continue the control operation after the occurrence of the software error, such as hardware diagnosis and correction of various setting values. To secure time for.

  In the period from time T1 to time T4, by reducing the control rate as described above, the dead time delay in the P control is increased and the phase margin is deteriorated. In order to compensate for this, in the period from time T1 to time T4, it is preferable to lower the proportional gain coefficient Kp than in the normal time.

  If the internal data of the integrator 2021 and the differentiator 2022 can be recovered, the controller 202 switches the control processing after the time T2 from P control to PID control in the same way as the first processing method, and before the soft error occurs. Return to the control state. As described in the first processing method, the internal data of the differentiator 2022 is recovered during the period from the time T1 to the time T3, and the differential value obtained by the differentiator 2022 is obtained during the period from the time T3 to the time T2. You may make it perform P control (PD control) which considered.

  In the third processing method shown in FIG. 5, when a soft error occurs at time T0, the controller 202 changes the control processing after time T1 from PID control to P control, as in the first and second processing methods. Switch to. After that, when the internal data of the integrator 2021 and the differentiator 2022 are recovered by time T5, the controller 202 gradually shifts from P control to PID control in the period from time T5 to time T2. Specifically, the values of the proportional gain coefficient Kp, the integral gain coefficient Ki, and the differential gain coefficient Kd are gradually changed from the P control value to the PID control value. Thereby, when switching from P control to PID control, it is suppressed that a control value changes rapidly, and smooth control switching is performed. After the complete transition to PID control at time T2, the control state before the occurrence of the soft error is returned as in the first and second processing methods.

  In the third processing method, as described in the first processing method, it is preferable to recover the internal data of the differentiator 2022 during the period from time T1 to time T3. In this case, as in the first and second processing methods, in the period of time T5 when the internal data of the integrator 2021 is recovered from the time T3 and the transition to PID control is started, it is obtained by the differentiator 2022. You may make it perform P control (PD control) which considered the differential value.

  FIG. 6 is a diagram showing changes in the proportional gain coefficient Kp, the integral gain coefficient Ki, and the differential gain coefficient Kd in the third processing method. As shown in FIG. 6, the controller 202 switches from PID control to P control by increasing the proportional gain coefficient Kp and setting the integral gain coefficient Ki and the differential gain coefficient Kd to 0 at time T1. Thereafter, when the internal data of the differentiator 2022 is recovered, PD control is performed by gradually increasing the differential gain coefficient Kd during the period from time T3 to time T5. At this time, the differential gain coefficient Kd may remain 0 and the P control may be continued.

  When the internal data of the integrator 2021 is recovered, the controller 202 gradually decreases the proportional gain coefficient Kp and gradually increases the integral gain coefficient Ki and the differential gain coefficient Kd during the period from time T5 to time T2. Thus, the P control (PD control) is gradually shifted to the PID control. When the proportional gain coefficient Kp, the integral gain coefficient Ki, and the differential gain coefficient Kd return to the values before the occurrence of the soft error at time T2, the controller 202 performs PID control in the subsequent period.

  FIG. 7 is a diagram showing a change in internal data of the integrator 2021 and the differentiator 2022 in the third processing method. As shown in FIG. 7, when a soft error occurs at time T0, the internal data of the integrator 2021 and the differentiator 2022 become indefinite. Thereafter, by performing reset processing, internal data of the integrator 2021 and the differentiator 2022 is reset to 0 at time T1.

  When the reset of the differentiator 2022 is released at time T6, the convergence operation of the differentiator 2022 is started, and the internal data of the differentiator 2022 is gradually recovered from zero. Subsequently, when the reset of the integrator 2021 is released at time T3, the convergence operation of the integrator 2021 is started, and the internal data of the integrator 2021 is gradually recovered from zero.

  As described above, in the third processing method, after the soft error occurs, the internal data of the integrator 2021 and the differentiator 2022 are reset and switched to the P operation, and then the proportional gain coefficient Kp, the integral gain coefficient Ki, and the differential gain are changed. The PID control is restarted by gradually changing the coefficient Kd. As a result, the internal data of the integrator 2021 is compared with the conventional windup countermeasure method in PID control in which the convergence prediction value of integration is used, or when the output is excessive, the integration operation is temporarily stopped. It can be quickly converged.

  FIG. 8 is a diagram showing how the output value changes when the control target value is converted into a rectangular wave shape. In this case, as shown in FIG. 8, when a soft error occurs at time T0, the steady-state deviation between the target value and the output value is higher than normal during the period from switching to P control to returning to PID control. Will also expand. The magnitude of the steady deviation at this time is expressed as 1 / (1 + Kp) using the proportional gain coefficient Kp because the control system of the P control corresponds to a 0-type control system in which the number of integrators is zero. Can do. On the other hand, the steady-state deviation is substantially zero before the soft error occurs and in the PID control period after the internal data recovery of the integrator 2022.

  FIG. 9 is a diagram showing how the output value changes when the control target value changes in the lamp input state. In FIG. 9, a waveform 91 shows an output value in the case of a type 1 control system in which one integrator 2021 is provided in the controller 202 as shown in FIG. The output value in the case of a type 2 control system in which two integrators 2021 are provided in the unit 202 is shown. Further, a waveform 90 shown as a reference in FIG. 9 indicates an output value in a zero-type control system in which the number of integrators is zero.

  When the control target value changes in the lamp input state, as shown in FIG. 9, if a soft error occurs at time T0, the target value and the output value are changed during the period from switching to P control until returning to PID control. The steady-state deviation during the period gradually increases. The slope of the output value at this time is equal to that of the 0-type control system in both the 1-type control system and the 2-type control system. On the other hand, in the PID control period before the occurrence of the soft error and after the internal data recovery of the integrator 2022, the steady-state deviation is a constant value in the case of the type 1 control system, and is substantially 0 in the case of the type 2 control system.

  According to the 1st Embodiment of this invention demonstrated above, there exist the following effects.

(1) The controller 202 functioning as a feedback control device is a control value for controlling the output value of the electric power steering device based on the d-axis error signal δId and the q-axis error signal δIq that are input values. An axis target voltage Vd and a q-axis target voltage Vq are determined. In this process, the controller 202 controls the d-axis target voltages Vd and q, which are control values, by PID control using the current input value and the past input value in a period before the time T0 when the soft error occurs. An axis target voltage Vq is determined. On the other hand, in the period from time T1 to time T2 (or time T3) after the soft error occurs, the controller 202 uses the current input value and the control value by the P control that does not use the past input value. A certain d-axis target voltage Vd and q-axis target voltage Vq are determined. Further, PID control is resumed in a period after time T2 after this period has elapsed. Since it did in this way, control operation can be continued safely also when a soft error occurs.

(2) In the PID control, the controller 202 uses the adder 2026 to multiply the current input value by a predetermined proportional gain coefficient Kp, and the integration value and differentiator by the integrator 2021 based on the past input value. The d-axis target voltage Vd and the q-axis target voltage Vq, which are control values, are determined by adding the calculated values obtained by multiplying the differential value obtained by 2022 by a predetermined integral gain coefficient Ki and a differential gain coefficient Kd, respectively. On the other hand, in the P control, the integral gain coefficient Ki and the differential gain coefficient Kd are set to 0 so that the integral value and the derivative value based on the past input values are not used. Since it did in this way, internal data of integrator 2021 and differentiator 2022 can be recovered, performing P control.

(3) In the third processing method shown in FIG. 5, in the period from time T5 to time T2, the controller 202 gradually increases the integral gain coefficient Ki and the differential gain coefficient Kd from 0 to the original value, respectively. Let Since it did in this way, when switching from P control to PID control, it can suppress that a control value changes rapidly, and can implement | achieve smooth switching of control.

(4) In the second processing method shown in FIG. 4, the controller 202 makes the P control cycle longer than the PID control cycle in the period from time T1 to time T4. Since it did in this way, processing time for the process required in order to continue control operation after soft error generation | occurrence | production can be ensured, performing P control.

(5) After a soft error occurs at time T0, the controller 202 resets an integral value and a derivative value based on past input values used in PID control. Since it did in this way, normal PID control can be performed after reset, even when an integral value or a differential value becomes an abnormal value due to a soft error.

(6) As shown in FIG. 8 and FIG. 9, the steady-state deviation between the target value and the output value is a predetermined value during the period in which the PID control is executed before the soft error occurs at time T0. On the other hand, in the period in which the P control is executed after the soft error occurs, the steady deviation between the target value and the output value becomes larger than the above value. As described above, during the period from when the soft error occurs until the internal data of the integrator 2021 and the differentiator 2022 is recovered, the steady-state deviation is larger than normal, but the control operation can be continued. it can.

(Second Embodiment)
FIG. 10 is a diagram showing an overall configuration of an electric power steering apparatus according to the second embodiment of the present invention. In FIG. 10, the difference from the electric power steering apparatus according to the first embodiment shown in FIG. 1A is that the control system 200 has a majority decision with magnetic pole position storage units 207-1, 207-2, and 207-3. And a portion 208.

  The magnetic pole position storage units 207-1, 207-2, and 207-3 are parts that function as a storage device for storing the magnetic pole positions of the motor 2 in a redundant manner. Information on the magnetic pole position output from a magnetic pole position sensor (not shown) provided in the motor 2 is input and stored in each of the magnetic pole position storage units 207-1 to 207-3.

  The majority decision unit 208 reads out the magnetic pole positions stored in the magnetic pole position storage units 207-1 to 207-3, takes these majority decisions, and outputs the result to the two-phase / three-phase conversion unit 203. Thus, even when an incorrect magnetic pole position is stored in any of the magnetic pole position storage units 207-1 to 207-3 due to a soft error or the like, a correct magnetic pole position can be obtained.

  Generally, brushless motors are frequently used in electric power steering devices. When a brushless motor is used as the motor 2, one of the points of concern when the control system 200 controls the motor 2 is that the control cannot follow the change in the magnetic pole position of the motor 2 and step out. . In particular, when a resolver is used for the magnetic pole position sensor, it is necessary to calculate the magnetic pole position from the sensor signal, so it takes time to obtain the magnetic pole position. Therefore, when the operation of the control system 200 is temporarily stopped and restarted due to a soft error, the magnetic pole position cannot be obtained in a timely manner, and there is a risk of step-out.

  Therefore, in the electric power steering apparatus of the present embodiment, the magnetic pole position information is redundantly stored in the magnetic pole position storage units 207-1 to 207-3 with the configuration shown in FIG. Thereby, even when a soft error occurs, step-out due to a time delay for calculating the magnetic pole position can be prevented, and the control operation can be continued.

  For example, if the control cycle of the motor 2 by the control system 200 is about 100 μs to 1 ms and it is necessary to perform control about 10 times to converge the integrator 2021, the internal data of the integrator 2021 is recovered. Takes about 1 ms to 10 ms. As described in the first embodiment, the present invention prevents the motor control from being stepped out by continuing the control operation while performing the P operation during this time. Further, when the resolver is used for the magnetic pole position sensor as described above, for example, if the excitation signal frequency is 10 kHz, detection of the magnetic pole position takes at least one period of the excitation signal, that is, about 100 μs. . In the present embodiment, the magnetic pole position information is made redundant and stored in the magnetic pole position storage units 207-1 to 207-3, so that this time can be shortened. It should be noted that for the return of the variables necessary for the operation of the operating system, when the clock frequency of the processor in the control system 200 is several hundred MHz, for example, a processing time of about several hundred to several thousand steps, that is, a time of about 1 to 10 μs. It takes. However, since this time is sufficiently shorter than the detection time of the magnetic pole position, there is no particular problem when the motor control is continued after the soft error occurs.

  According to the second embodiment of the present invention described above, the electric power steering apparatus further includes magnetic pole position storage units 207-1 to 207-3 as storage apparatuses that store the magnetic pole positions of the motor 2 in a redundant manner. . Since it did in this way, the step-out at the time of continuing motor control after soft error generation | occurrence | production can be suppressed effectively.

(Third embodiment)
In each embodiment described above, PID control is performed as feedback control using the current input value and the past input value before the soft error occurs and after the internal data recovery of the integrator 2021. An example of a feedback control device that performs P control as feedback control using the current input value and not using the past input value during the period from when the internal data of the integrator 2021 is restored to when the internal data of the integrator 2021 is recovered has been described. However, the present invention is not limited to this, and can also be applied to feedback control devices that perform various feedback controls, such as feedback control based on modern control theory. Hereinafter, as a third embodiment of the present invention, an example in which the present invention is applied to a feedback control apparatus that performs feedback control based on modern control theory will be described.

  In general, in feedback control based on modern control theory, the relation between the state of the system to be controlled and the input is represented by a state equation, and the control value for the control deviation input is determined based on the system state obtained from this state equation. Thus, feedback control is performed. For example, when an input value is represented by u (t) and a system state is represented by x (t), a state equation representing these relations is represented by the following equation (2). However, in Formula (2), A and B are predetermined coefficient matrices, respectively.

  When the control value output here is expressed as y (t), the control value y (t) is obtained from the input value u (t) and the system state x (t) as shown in the following equation (3). It is done. However, in Formula (3), C and D are predetermined coefficient matrices, respectively.

  In the above equation (3), the system state x (t) is history data that is updated by repeating the operation of the equation (2).

  FIG. 11 shows a controller 202 according to the third embodiment of the present invention in the case where feedback control based on the above-described modern control theory, that is, control based on a past system state (hereinafter referred to as “state control”) is performed. It is a control block diagram of. As illustrated in FIG. 11, the controller 202 that performs state control in the present embodiment includes a state generation unit 202a and an output signal generation unit 202b.

  The following equation (4) is derived by integrating both sides of the aforementioned equation (2).

  In the controller 202 of the present embodiment shown in FIG. 11, the state generator 202a is based on the equations (2) and (4), and integrators 2121 and amplifiers 2122 and 2123 corresponding to the coefficient matrices A and B, respectively. And an adder 2126. On the other hand, the output signal generation unit 202b is configured by control blocks of amplifiers 2124 and 2125 and adders 2127 respectively corresponding to the coefficient matrices C and D, based on Expression (3).

  The coefficient matrices A, B, C, and D represent matrix operations, respectively. For example, when the system state x (t) is n-dimensional, the input value u (t) is m-dimensional, and the control value y (t) output from the controller 202 is one-dimensional, the coefficient matrices A, B, C , D are represented by determinants of A: n × n, B: n × m, C: 1 × n, and D: 1 × m, respectively.

  In the present embodiment, as described in the first embodiment, when the comparator 211 makes an abnormality determination and a failure (SEU) occurrence is detected, the controller 202 temporarily changes the value of the coefficient matrix C. 0. At this time, the control value y (t) output from the controller 202 is expressed by the following equation (5) by substituting C = 0 in the above equation (3).

  According to the above equation (5), by setting C = 0 in the controller 202, it is possible to shift from the above-described state control to control that does not use the past system state (hereinafter referred to as “non-state control”). I understand. During the execution of the non-state control, the controller 202 continues the operation of the motor 2 to be controlled, while the system state x (t) stored in the integrator 2121 destroyed by the failure (SEU). Allow the value to converge and recover. Thereafter, the controller 202 can resume the state control.

  In the present embodiment, the input value u (t) to the controller 202 and the control value y (t) output from the controller 202 are both values represented by multiple variables. Here, if U (t) = (δIq, δId) and y (t) = (Vq, Vd), the present invention can be applied to the electric power steering apparatus of FIG. 1A described in the first embodiment.

  FIG. 12 is a diagram showing changes in the coefficient matrices A, B, C, and D in the third embodiment of the present invention. The coefficient matrices A, B, C, and D are each represented as a determinant composed of a plurality of matrix elements. In FIG. 12, each coefficient is shown in order to make it easy to understand how these changes occur. The size of the matrix is represented by the size of an arbitrary element in each determinant.

  As shown in FIG. 12, the controller 202 switches from state control to non-state control by raising the coefficient matrix D and setting the coefficient matrix C to 0 at time T1. After that, when the internal data of the integrator 2121 is recovered, the controller 202 gradually decreases the coefficient matrix D and gradually increases the coefficient matrix C in the period from time T5 to time T2, thereby performing non-state control. Gradually shift from state to state control. When the coefficient matrices C and D return to the values before the failure at time T2, the controller 202 performs state control in the subsequent period.

  Here, as is clear from Expression (5), if C = 0, the value of the system state x (t) does not affect the value of the control value y (t) output from the controller 202. Therefore, it is not necessary to change the values of the coefficient matrices A and B at the time of non-state control, but rather the convergence and recovery of the system state x (t) can be promoted by not changing the values as shown in FIG. Is desirable.

  FIG. 13 is a diagram showing a state of change of internal data of the integrator 2121 according to the third embodiment of the present invention. Although the internal state of the integrator 2121 is represented by multiple variables, in FIG. 13, in order to make the state of the change of the internal state of the integrator 2121 easy to understand, the value of the internal data value of the integrator 2121 is large. This is represented by the size of an arbitrary element (variable) among multiple variables.

  As shown in FIG. 13, when a soft error occurs at time T0, the internal data of the integrator 2121 becomes indefinite. Thereafter, by performing a reset process, the internal data of the integrator 2121 is reset to 0 at time T1. When the reset of the integrator 2121 is released at time T3, the convergence operation of the integrator 2121 is started, and the internal data of the integrator 2121 is gradually recovered from 0. After the integrator 2121 is converged in this way, after time T2, the coefficient matrices C and D are set to the same values as before the failure occurrence, and the original state control is restored.

  As described above, according to the present embodiment, in the feedback control based on the modern control theory, even when the history data of the system state x (t) necessary for the state control is destroyed due to the failure (SEU), the non-state control The control operation can be continued. Further, the history data of the system state x (t) can be converged and recovered while continuing the control operation in the non-state control. Therefore, it is possible to resume the state control without a control step.

  In modern control theory, for example, an observer is used to estimate a state quantity inside the system. A Kalman filter is often used to remove the influence of disturbance. In the Kalman filter, the estimated value of the system state can be expressed as history data expressed in the form of a linear combination using all of the observation values observed so far. When the system receives a control input, an estimated value of the system state is obtained including history data of the input value.

  Moreover, although each embodiment demonstrated above demonstrated the case where the one inverter 100 was mounted in the electric power steering apparatus, the some inverter may be mounted. In this case, by driving a plurality of motors or windings using a plurality of inverters, it is possible to make the inverters, motors, and windings redundant, so that the operation can be continued even if these failures occur.

  In each of the embodiments described above, an example of a feedback control device that performs feedback control in the electric power steering device has been described. However, the present invention is applied to devices that are used in other devices and systems as long as the feedback control is performed. May be.

  Each embodiment and various modifications described above are merely examples, and the present invention is not limited to these contents as long as the features of the invention are not impaired. The present invention is not limited to the above-described embodiment, and various modifications can be made without departing from the spirit of the present invention.

2: motor, 100: inverter, 200: control system, 201: current command value calculation unit, 202: controller, 203: two-phase / three-phase conversion unit, 204: duty calculation unit, 205: three-phase / two-phase conversion unit, 207-1, 207-2, 207-3: magnetic pole position storage unit, 208: majority decision unit, 2021: integrator, 2022: differentiator, 2023, 2024, 2025: amplifier, 2026: adder

Claims (10)

A feedback control device that determines a control value for controlling the output value based on an input value corresponding to a difference between a target value and an output value,
In the first period before the soft error occurs, the control value is determined by the first feedback control using the current input value and the past input value,
In a predetermined second period after the soft error occurs, the control value is determined by second feedback control using a current input value and not using a past input value,
A feedback control device that resumes the first feedback control in a third period after the second period has elapsed. The feedback control device according to claim 1,
In the first feedback control, a calculated value obtained by multiplying the current input value by a predetermined first coefficient and a calculated value obtained by multiplying a value based on the past input value by a predetermined second coefficient are totaled. To determine the control value,
In the second feedback control, the second coefficient is set to 0 so that a value based on the past input value is not used. The feedback control device according to claim 2,
A feedback control device that gradually increases the second coefficient from 0 to an original value in a predetermined fourth period provided between the second period and the third period. The feedback control device according to claim 1,
A feedback control device that makes a period of the second feedback control longer than a period of the first feedback control in a predetermined fifth period of the second period. The feedback control device according to claim 1,
In the first period, PID control is performed as the first feedback control,
In the second period, a feedback control apparatus that performs P control as the second feedback control. The feedback control device according to claim 1,
In the first period, a feedback control device that performs feedback control based on a state equation based on modern control theory as the first feedback control and the second feedback control. The feedback control device according to claim 1,
A feedback control device that resets a value based on the past input value used in the first feedback control after the soft error occurs. A feedback control device that determines a control value for controlling the output value based on an input value corresponding to a difference between a target value and an output value,
In the first period before the soft error occurs, a steady deviation between the target value and the output value is a predetermined first value,
A feedback control device, wherein a steady-state deviation between the target value and the output value is larger than the first value in a predetermined second period after the soft error occurs. A feedback control device according to any one of claims 1 to 8,
A motor that generates assist torque for the steering mechanism;
An electric power steering device comprising: an inverter that drives the motor based on a control value determined by the feedback control device. The electric power steering apparatus according to claim 9,
An electric power steering device further comprising a storage device for storing the magnetic pole position of the motor in a redundant manner.

Images